Hi, I am training some neural networks using Flux.jl.
I noticed that the output of display is much nicer than the output of print and show.
I am making a file for evaluating which architectures I think look promising. For this I would like to take the output that looks similar to the one from display:
julia> print(NN)
Chain(Conv((2, 2), 2 => 64, relu), Conv((2, 2), 64 => 16, relu), flatten, Dense(2704 => 256, relu), Dense(256 => 128, relu), Dense(128 => 3))
julia> show(NN)
Chain(Conv((2, 2), 2 => 64, relu), Conv((2, 2), 64 => 16, relu), flatten, Dense(2704 => 256, relu), Dense(256 => 128, relu), Dense(128 => 3))
julia> display(NN)
Chain(
Conv((2, 2), 2 => 64, relu), # 576 parameters
Conv((2, 2), 64 => 16, relu), # 4_112 parameters
Flux.flatten,
Dense(2704 => 256, relu), # 692_480 parameters
Dense(256 => 128, relu), # 32_896 parameters
Dense(128 => 3), # 387 parameters
) # Total: 10 arrays, 730_451 parameters, 2.787 MiB.
In the method description of display it does not seem to me like it is possible to give an output stream to display.
Is there maybe a different method that I can call that produces a similar output? Or a way to redirect its output?